Developer Insights #18 - Graphics of Early Access KSP2

[RocketRockington]

If PQS is such garbage that KSP2's 'PQS+' system that they optimized and reworked and 'pushed to its limits' is limited to taking 24ms on a 1060 then how come modders (Parrallax in particular) - who have work around both the limitations of the old PQS system - have managed to make the old PQS system on KSP look better than KSP2 & be performant enough to run on older hardware.

Also - why wait till now to implement CBT (unfortunate acronym there).  Even if we say KSP2 got a restart in 2020 - shouldn't they have learned from both KSP1 and KSP2 that PQS wasn't going to work out for them?

Some of this is really good detail and I appreciate the engineer for writing it.  Some of it feels like the standard KSP2 spin and nonsense to make it seem like the emperor is still wearing underwear even after they admit he's mostly naked.  Guys, you've released the game.  We know what it looks like.  Please stop with the marketting spin.

Edited March 10, 2023 by RocketRockington

[TackleMcClean]

Thanks for the update, to me it sounds like the right decision and we'll hopefully see better and more performant graphics.

[Haustvindr]

20 minutes ago, PDCWolf said:

Your initial paragraph makes sense, until you add in the fact that PQS being limited was something we knew at least 10 years ago. Whilst hardware has improved significantly, terrain remains the biggest source of frame delay in KSP1 (if you don't push partcounts). This has been known for a long time, as the last update to PQS was on 0.20, in the form of "streamlining", and even then it was still easily noticeable that terrain and the PQS system in general was a problem. The appearance of mods like Kopernicus (and back then some non Kopernicus planet mods) made this even more painfully evident.

Oh I know, I really know the pain of being too close to Kerbin. Thing is, we don't know what kind of modifications and improvements they did to the original PQS, so we don't know how much improvement there is. Through their words, it seems they deemed them sufficient enough even when accounting for the current not-fully-optimized state, so it seems their revamping included bettering KSP1 system all the way around and not just applying things on top.

To me, it reads as if CBT was going to come anyways, but probably much later on, probably well into full release. Thing is... it seems they still want to apply more things on top of their PQS+, but after diving into these days feedback data from the people playing, they decided against it and to spend the time to bring CBT now.

Either way, as I said it's a guess, no more than an opinion. We won't really know, but it's fun to speculate!

Edit: And now I'm actually curious, about the neat new terrain with all those details we've seen in the works, was that still PQS+ or was that already experimenting with CBT?

Edited March 10, 2023 by Haustvindr

[JoeSchmuckatelli]

20 minutes ago, Stoup said:

As happy as I am for the news... it seems like it'd be logical to reconsider if there's any other areas of the game that might be due for a proper refactor like this instead of trying to trudge along old code. I'd hate to see us running into similar walls due to constraints placed by similarly neglected code in other areas of the game!

Checkout ShadowZone's recent interview.  Nate referenced this work - along with other, apparently parallel, efforts to improve the playability.

So - it's happening.

[The Aziz]

16 minutes ago, RocketRockington said:

have managed to make the old PQS system on KSP look better than KSP2

Mmmmmmmno, not really. KSP1 planets have a tiny fraction of the "quads", chunks, you name it, compared to what KSP2 has and no mod in history changed it. Just adding a ton of scatter didn't change the fact that the terrain was blocky and made of gigantic squares.

In KSP2 the squares are still there but are much much smaller, allowing for much more detailed terrain.

[Vanamonde]

Some comments removed. 

Please watch your language. 

Please stick to the topic. 

And if you say something in one post there's no need to say it again a couple of posts later. 

[RocketRockington]

1 hour ago, The Aziz said:

Mmmmmmmno, not really. KSP1 planets have a tiny fraction of the "quads", chunks, you name it, compared to what KSP2 has and no mod in history changed it. Just adding a ton of scatter didn't change the fact that the terrain was blocky and made of gigantic squares.

 

Guess which one of these screenshots is modded KSP1, and which is KSP2?  Maybe KSP2 has more quads in scene, or is running more lines of shader code, but subjectively - I know which game's appearance I prefer.  

But even just looking for tesselation density, and not admiring how much better modded KSP1's scatter and lighting and terrain textures are - look at the horizon lines.  I can see more visible poly corners on KSP2 than I can on modded KSP1.   So maybe... KSP2 doesn't actually do that many more poly's, or maybe they just use them very poorly.

 

Edited March 11, 2023 by RocketRockington

[mucco]

Very interesting read, thanks!

Have you been tracking performance for AMD GPU users as well? The very best visual part of the game - atmospheric planets - is completely borked for me and many other AMD GPU users. I had to turn Kerbin clouds to a minimum due to the horror, and Jool from up close looks like an eye with a blotchy blurred pupil surrounded by plain green. I haven't dared to go to Eve... That and the save corruption are the two main things preventing me from enjoying the game, otherwise it seems fun!

[RocketRockington]

Btw Chris, sorry you have to be exposed to the fan maelstrom.  Sucks to be on the job such a short time and have to deal with it.    I've been there.  At least none of it is your fault.  We're not reacting to your work, just KSP2's long history of overhyping and underdelivering,  

[Speeding Mullet]

29 minutes ago, RocketRockington said:

Guess which one of these screenshots is modded KSP1, and which is KSP2?

It's obvious, but also for me not a very fair comparison.  One screenshot is of a game that has just come out in early access, and the other is of a game nigh on twelve years down the line, eight years after full release, and subject to the best the (very skilled) community can do.  This shot is also KSP2 (unmodded) and looks fantastic:

Looking back at some of my (not even old) screenshots of modded KSP1 this looks superior in many ways, so it appears it's really only in the last few years that good has become great for KSP1. I would say it's a very safe bet that KSP2 modded (and probably stock) fidelity will far surpass modded KSP1 down the line, and it certainly won't take 8-10 years to achieve that as it has with KSP1.

On a more general note and not in reply to your comment - I was mega hyped for KSP2, yet it hasn't (apart from the most glaring bugs which I'm confident will be patched in short order) disappointed me at all.  Maybe my throttle is just set to zero on launch or something.

SM

 

[RocketRockington]

1 hour ago, Speeding Mullet said:

It's obvious, but also for me not a very fair comparison.  One screenshot is of a game that has just come out in early access, and the other is of a game nigh on twelve years down the line, eight years after full release, and subject to the best the (very skilled) community can do.  This shot is also KSP2 (unmodded) and looks fantastic:

Your right, its not fair.  One game we're being asked to pay $50 to endure a crashy mess, one game is stable and was oh... I dunno... free on Epic a couple months ago.   Or you can buy it for $10.  Sorry as a consumer I don't have to listen to a company's excuses and give them the benefit of the doubt.   

More importantly, as it pertains to this devblog, they are specifically saying they pushed PQS 2.0 to the limit and it chugs and doesn't look as good as modded KSP using PQS 1.0.  (even your screenshot is not as pretty, though closer) .  So either PQS 2.0 is worse than 1.0, or its not being utilized as well as it could be.

So they want the community to wait even longer for the silver bullet fix, or just deal with uglier terrain on lower settings.  Hopefully it'll be worth the wait - KSP2 in its much delayed early access state hasn't been.

Edited March 11, 2023 by RocketRockington

[Nate Simpson]

I see a few people here interpreting Mortoc's post as an admission that we just took the terrain system from KSP1, slapped a new coat of paint on it, and called it PQS+. I'd like to clarify a few points here - again, from a non-specialist perspective, but based on what I witnessed as our terrain system evolved. While PQS+ does tessellate terrain in the same way that the original system did (that is to say, increasing the poly count when an observer/object gets close to the surface), it also does many, many things that the older system did not and is fundamentally a new system. Among the new requirements - it needed to support triplanar mapping, it needed to allow our terrain artists to intermix multiple distinct biomes in ways that looked good at all scales, it needed to support many octaves of detail so that artificial-looking patterns did not emerge at either high or low altitudes, it needed to support the addition of topology-defining decals for areas of high detail, and it required painstaking co-development with our environment artists to give them the tools to bring all of those things together in a way that looked natural and beautiful. I am actually quite proud of the progress made by our team on this system. I'm also very excited about ongoing performance improvements, as well as all the amazing new tech that Mortoc is developing.

PQS+ has performance challenges, but it has brought a huge number of new capabilities to KSP2 - namely, unlocking the vision of our creative team so that they could reimagine the Kerbolar System and craft the new star systems that will arrive over the course of Early Access. I think the new planets look stunning, and I think the terrain system almost does itself a disservice by doing so many formerly impossible things so well - for example, highly varied planetary surfaces betraying no tiling or biome transitions at high altitudes, or procedurally mixing different kinds of terrain so seamlessly that you forget how magical it is that nothing ever actually repeats. It provided us with the ability to be ambitious in our vision for KSP2, and I'm very thankful to the many people who worked very hard on the system to get it to where it is today.

That said, Mortoc is super rad and we're all pumped to see what comes next!

[Chilkoot]

4 hours ago, NoMrBond said:

What renderer is the game currently using?

SDRP

[didkodidko]

I cried, KSP really needs a modern terrain engine.

Hope it'll affect the physics interactions in a positive way.

 

Thank you 

[MrDiver]

11 minutes ago, Nate Simpson said:

I see a few people here interpreting Mortoc's post as an admission that we just took the terrain system from KSP1, slapped a new coat of paint on it, and called it PQS+. I'd like to clarify a few points here - again, from a non-specialist perspective, but based on what I witnessed as our terrain system evolved. While PQS+ does tessellate terrain in the same way that the original system did (that is to say, increasing the poly count when an observer/object gets close to the surface), it also does many, many things that the older system did not and is fundamentally a new system. Among the new requirements - it needed to support triplanar mapping, it needed to allow our terrain artists to intermix multiple distinct biomes in ways that looked good at all scales, it needed to support many octaves of detail so that artificial-looking patterns did not emerge at either high or low altitudes, it needed to support the addition of topology-defining decals for areas of high detail, and it required painstaking co-development with our environment artists to give them the tools to bring all of those things together in a way that looked natural and beautiful. I am actually quite proud of the progress made by our team on this system. I'm also very excited about ongoing performance improvements, as well as all the amazing new tech that Mortoc is developing.

PQS+ has performance challenges, but it has brought a huge number of new capabilities to KSP2 - namely, unlocking the vision of our creative team so that they could reimagine the Kerbolar System and craft the new star systems that will arrive over the course of Early Access. I think the new planets look stunning, and I think the terrain system almost does itself a disservice by doing so many formerly impossible things so well - for example, highly varied planetary surfaces betraying no tiling or biome transitions at high altitudes, or procedurally mixing different kinds of terrain so seamlessly that you forget how magical it is that nothing ever actually repeats. It provided us with the ability to be ambitious in our vision for KSP2, and I'm very thankful to the many people who worked very hard on the system to get it to where it is today.

That said, Mortoc is super rad and we're all pumped to see what comes next!

100% Agreed, Mortoc is doing Amazing work. I think the changes that will come in the future will make a big difference and also that the current system is also a HUGE improvement over KSP1. Even if it was based on KSP1 the new implementation just looks, feels and runs a lot better on average.

I am looking forward to more Dev Blogs in the future discussing these interesting Topics. Much love to Mortoc for taking the time to write this Amazing Dev Blog! <3

[ericye16]

Thanks so much for this update! As a software engineer not in games, I love reading about the challenges in making games. Loving KSP2 so far as well, keep up the good work!

[Haustvindr]

41 minutes ago, Nate Simpson said:

I see a few people here interpreting Mortoc's post as an admission that we just took the terrain system from KSP1, slapped a new coat of paint on it, and called it PQS+.

Oops, mea culpa, I guess. In my defense I can say that I thought about it as a complete renewal and not just a new coat.

41 minutes ago, Nate Simpson said:

I'm very thankful to the many people who worked very hard on the system to get it to where it is today.

Bringing new things never demerits the work done in the older ones, neither demerits the people who worked on them. Never, ever, and I'll volunteer to fight whoever says otherwise (but, uh, can we fight metaphorically?).

And people are enjoying a lot PQS+ judging all the screenshots, and will continue to do so for as long as you allow it!

Edited March 11, 2023 by Haustvindr

[tsaven]

Chiming in to give another vote of appreciation for the candidness of this post, and the technical explanations of what's going on.

But, uh . . . yeah, you might want to call it something besides CBT.  That acronym has certain associations when it comes to the internet.

[RocketRockington]

3 hours ago, Nate Simpson said:

I see a few people here interpreting Mortoc's post as an admission that we just took the terrain system from KSP1, slapped a new coat of paint on it, and called it PQS+. I'd like to clarify a few points here - again, from a non-specialist perspective, but based on what I witnessed as our terrain system evolved. 

I really suggest you consult with specialists before you make these posts then, because  you're saying things that are on the face of it, not true.

You've said PQS+ unlocks trilinear texturing.  That's... just wrong.  PQS is a subdivision system for generating tesselated surface detail from a height map.  Triplanar mapping is a texturing system that will texture a surface based on its world space coordinates.  Those are two different steps in a process.  One is not dependent on the other.  You can triplanar map a cube if you wanted to.  

PQS+ therefore didn't make this technique 'possible' - it wasn't impossible before.  The parrallax mod I mentioned does it already in KSP1. 

[snip]

Edited March 11, 2023 by Vanamonde

[MiffedStarfish]

1 hour ago, Nate Simpson said:

PQS+ has performance challenges, but it has brought a huge number of new capabilities to KSP2 - namely, unlocking the vision of our creative team so that they could reimagine the Kerbolar System and craft the new star systems that will arrive over the course of Early Access. I think the new planets look stunning, and I think the terrain system almost does itself a disservice by doing so many formerly impossible things so well - for example, highly varied planetary surfaces betraying no tiling or biome transitions at high altitudes, or procedurally mixing different kinds of terrain so seamlessly that you forget how magical it is that nothing ever actually repeats. It provided us with the ability to be ambitious in our vision for KSP2, and I'm very thankful to the many people who worked very hard on the system to get it to where it is today.

But the planets just don't look good outside of the map view? Like what are you even talking about? I can look at Kerbin and see jagged square tiles of terrain where it intersects the ocean at basically any range, nearby hills appear as perfect triangles because the quality vs distance is worse than KSP. That's not an improvement. I don't care if it apparently never looks like it repeats when the ground looks magnitudes worse overall, the scatter assets seem to be UV unwrapped more incorrectly than not, the mountain ranges have shrunk ridiculously and are completely unimpressive and unexciting now?

Like usual the game you talk about and the reality of the game in our hands are completely different things.

Edited March 11, 2023 by MiffedStarfish

[IronGremlin]

21 minutes ago, RocketRockington said:

PQS+ therefore didn't make this technique 'possible' - it wasn't impossible before.  The parrallax mod I mentioned does it already in KSP1. 
 

 

I was unaware parallax was a part of stock KSPs terrain rendering system?

 

I get where you're coming from as an engineer, but it is actually totally legit to overhaul and expand a system with new capabilities that the old system did not have any then make the claim that the old systems did not have those capabilities - that is a perfectly logically consistent statement that is in no way false.

[Speeding Mullet]

1 hour ago, RocketRockington said:

one game is stable

After a long journey to get it to where it is today.  I'm not going to get into the pricing argument as people see value very differently depending on their view point.

1 hour ago, RocketRockington said:

So either PQS 2.0 is worse than 1.0, or its not being utilized as well as it could be.

This has been covered by Nate, and Mortoc said that PQS+ has served them well but CBT will give better results in terms of performance and visual quality.  I think it looks great already, so its just another thing that get's me excited about the future of the game. Hopefully you can enjoy the game more in the near future.

One interesting thing I learned about sayings with Emperor in (although I absolutely don't share your opinion on that) is that there's a big difference between "The Emperor Wears No Clothes" and "The Emperor's New Clothes".  Made for some interesting reading!

SM

[KingWilliamB]

What confuses me about this is why wouldn't they have started with CBT in the first place? The support for it in unity wasn't added that long ago from what I've heard.

[HopelessSoap]

18 minutes ago, MiffedStarfish said:

But the planets just don't look good outside of the map view? Like what are you even talking about? I can look at Kerbin and see jagged square tiles of terrain where it intersects the ocean at basically any range, nearby hills appear as perfect triangles because the quality vs distance is worse than KSP. That's not an improvement. I don't care if it apparently never looks like it repeats when the ground looks magnitudes worse overall, the scatter assets seem to be UV unwrapped more incorrectly than not, the mountain ranges have shrunk ridiculously and are completely unimpressive and unexciting now?

Like usual the game you talk about and the reality of the game in our hands are completely different things.

[RocketRockington]

34 minutes ago, IronGremlin said:

 

I was unaware parallax was a part of stock KSPs terrain rendering system?

 

I get where you're coming from as an engineer, but it is actually totally legit to overhaul and expand a system with new capabilities that the old system did not have any then make the claim that the old systems did not have those capabilities - that is a perfectly logically consistent statement that is in no way false.

I didn't say it was.  I'm just saying modders added triplanar mapping to KSP. Nothing about the old PQS precluded it - which is what Nate told us 

Claiming that it was 'impossible' to support a feature under the old system when modders, who don't even have full access to the code, added, is a falsehood.  Especially because his explanation makes no sense to someone who knows the jargon he's speaking about.

Edited March 11, 2023 by RocketRockington